Google lets creators strip visible watermarks from AI-generated content

Google announced Friday that it will let users strip visible watermarks from material created through its artificial intelligence tools—spanning images, video, and music. The change addresses a practical friction point: visible markers often diminish the usability of AI outputs for professional and creative projects.

The invisible SynthID watermark and C2PA metadata will remain in place regardless of the user's choice on visible marking.

How it will work#

Josh Woodward, Google's VP for Gemini, indicated the toggle will apply to the Nano Banana, Omni, and Lyria generation models. Users will find the visibility control in Gemini's settings and the company's AI video editor Flow, with Search integration coming later. The rollout begins within days; once live, toggling watermarks on and off will happen via Settings > Media Watermark.

Balancing disclosure with utility#

Google framed the shift as a middle ground. According to Woodward, "We're striking a balance here between creative control and safety: while the visible watermarks are now optional, invisible SynthID watermarks and C2PA metadata are still being used for transparency. So you can still use Gemini or Search to see if an image was AI-generated."

The hidden markers preserve a means of verification without compromising visual appeal or workflow efficiency.

Developer tools#

Alongside the watermark change, Google is open-sourcing Credentio, a library that enables developers to embed local validation of media credentials directly in their applications.

The move follows similar moves by AI labs weighing transparency against usability—notably Anthropic's recent decision to watermark Claude-generated text and files to meet EU compliance requirements, a decision that drew pushback from some users.